Agent Skill Authoring

Use this capability whenever you create, refine, split, consolidate, rename, or audit an agent skill under the host project's skill root. It is what turns a durable convention into a well-formed, discoverable skill and keeps the skill tree coherent as it grows.

Skills authored here follow the agentskills.io format. For the host project's active skill inventory and topic-to-skill routing, defer to each skill's own description discovery metadata and the directory listing under the skill root; where a host also maintains a written index (e.g. AGENTS.md), keep it in sync too.

Guidelines:

• MUST run the bundled validator that owns what a change touched — scripts/check-skill-frontmatter.mjs after editing frontmatter, scripts/check-skill-body.mjs after editing prose, scripts/check-skill-references.mjs after adding, moving, or renaming a reference file — and all three when unsure; together they are the enforcement path for the frontmatter, naming, discovery length-cap, and reference-linkage rules this skill states nowhere else (see audit-checklist.md).
• SHOULD propose or implement a skill update when any task exposes a reusable convention, outdated guidance, a recurring review issue, or a missing project rule — skill maintenance happens when work reveals durable learning, not after every narrow fix.
• SHOULD skip skill maintenance when the work produced no generalizable learning, and state that it was skipped.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.

Scoping and MECE

See scoping-and-mece.md for: